refactor(question_gen): adapt generator/gates/store signatures for strategy

- generator_v2: _load_prompt_template takes template_name str instead of
  QuestionFamilySpec; _build_v2_prompt takes prompt_template + strategy_name
  + sub_pattern_instruction; generate_one_v2 takes discrete params
  (prompt_template, strategy_name, skill_target, sub_pattern_instruction)
- gates: _gate_leak_test and run_gates take leak_probe_template str
  instead of QuestionFamilySpec
- run_store: add sub_pattern column to DDL + idempotent migration;
  record_item accepts optional sub_pattern param
- Remove QuestionFamilySpec imports from generator_v2 and gates modules
- Update test call sites accordingly

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tests/unit/test_gates.py

@@ -7,7 +7,6 @@ from typing import Any
 
 import pytest
 
-from app.question_gen.families import RETRIEVAL_FAMILY
 from app.question_gen.gates import (
     CandidateQuestion,
     GateVerdict,
@@ -73,7 +72,22 @@ class MockLLM:
         self._call_count += 1
         if idx < len(self._responses):
             return self._responses[idx]
-        # 默认返回 PASS
+        return _make_llm_response("pass", "default")
+
+    async def chat_with_images(
+        self,
+        messages: list[dict[str, Any]],
+        images: list[Any],
+        *,
+        session_id: str | None = None,
+        parent_call_id: str | None = None,
+    ) -> LLMResponse:
+        """记录 VLM 调用并返回预设响应（与 chat 共享计数器）。"""
+        self.calls.append({"messages": messages, "images": images, "session_id": session_id})
+        idx = self._call_count
+        self._call_count += 1
+        if idx < len(self._responses):
+            return self._responses[idx]
         return _make_llm_response("pass", "default")
 
 
@@ -254,11 +268,13 @@ class TestGateLeakTest:
     """leak_test 门：按家族模板执行泄漏探测。"""
 
     @pytest.mark.asyncio()
-    async def test_per_family_template(self) -> None:
-        """使用家族特定的 probe_template 调用 LLM。"""
+    async def test_per_strategy_template(self) -> None:
+        """使用策略特定的 probe_template 调用 LLM。"""
         llm = MockLLM([_make_llm_response("pass", "no shortcut detected")])
         candidate = _make_candidate()
-        result = await _gate_leak_test(candidate, RETRIEVAL_FAMILY, llm, session_id="test-session")
+        result = await _gate_leak_test(
+            candidate, "gate_leak_retrieval.md", llm, session_id="test-session"
+        )
         assert result.verdict == GateVerdict.PASS
         # 验证 session_id 被正确传递
         assert llm.calls[0]["session_id"] == "test-session"
@@ -284,7 +300,7 @@ class TestRunGates:
             candidate=candidate,
             tree=tree,
             llm=llm,
-            family_spec=RETRIEVAL_FAMILY,
+            leak_probe_template="gate_leak_retrieval.md",
             postprocess=postprocess,
             session_id="test-session",
         )
@@ -303,7 +319,7 @@ class TestRunGates:
             candidate=candidate,
             tree=tree,
             llm=llm,
-            family_spec=RETRIEVAL_FAMILY,
+            leak_probe_template="gate_leak_retrieval.md",
             postprocess=postprocess,
             session_id="test-session",
         )